6 Questions to Ask Before You Invest in Conversational AI

Great news: you’ve decided to leap into the world of automation, or maybe you’ve done some limited pilot projects, proved a point or two, and need to roll out. Your next step will be to assess chatbot technologies and automation platforms to ensure you find the right capabilities for your business and contact center needs.

Your organization is counting on you to find the right technology to partner with, and this is a long-term investment that can be scaled across your business, making it critical to find the right solution. But Conversational AI vendors vary greatly, so how do you instill confidence in your evaluation? What’s a good starting point in your evaluation journey?

Here are six important questions to ask Conversational AI vendors:

1. Is this a low-code solution?

When it comes to creating virtual agents, the best people to do this are often non-technical employees who work on the front line with customers. They know the questions that customers generally ask and what replies they need to be given to answer their queries as quickly as possible. Your ops people may not be technical experts, but they are masters at customer service and maximizing customer satisfaction – which is exactly what you want your bots to do. A low-code platform with a visual developer environment enables operations and non-technical people to create bots without having to code.

In regard to training bots, it should be possible for non-technical people to build agents entirely from scratch, and even add new integrations to your existing systems, without having to write a line of code. One reason for this is that it frees up time that your valuable (and probably expensive) IT people could be spending on other things. Another reason is that conversations, interactions, and customer journeys are better built by operational and front-line staff who understand your brand and customers and what the customer experience should look like.

2. Is your solution open source, proprietary, or a mix of both?

Open-source software can be great, as you can potentially make it do exactly what you want, but only if you have a serious level of technical expertise and coding talent in-house, as well as the time to build your own solution from the ground up. The sweet spot is a solution built on open source standards but supported by a strong vendor. This generally gives you a platform that is almost infinitely customized, easy to integrate with other software, and can be extended to offer whatever functionality you need.

3. How does your solution handle fail-over to a live agent?

Most chatbots and automated systems are built to recognize user inputs and then respond with pre-programmed responses. When a user’s intent is identified as falling outside of those the bot has been built to handle, it needs to fail gracefully. It’s important to understand how your proposed solution handles this, how it integrates with your contact center’s voice or live chat platform, and what level of control you have over how and when the handover happens.

4. Can we build end-to-end customer journeys across channels?

Most chatbots are single-purpose and built to work on a single channel. It should be possible, however, in a modern and versatile automation platform, to chain bots together in order to create entire customer journeys. Using nodes and endpoints (coding and API terms which allow systems to be connected together) it should be possible to create branching conversation trees that even take in multiple channels, such as passing the customer between voice and chat or making an automated SMS follow up to a chat to check the customer’s query was correctly handled.

5. What software do we need in our contact center to use your platform?

The ideal answer to this question is: Whatever you already have. In addition to integration with channels (see the next question), integrations with the contact center and back-office systems are of course required for virtual agents to initiate activities like handing over to a live agent, scheduling a callback or follow-up SMS, and triggering a business process such as a dispatch, payment, or refund. When assessing a platform it’s useful to understand how many pre-built integrations it includes, and what tools and APIs are included so that you can build new ones.

Your potential vendor should support as many of the popular third-party channels as possible and also a healthy number of channel aggregators, right out of the box. Channel aggregators, such as DialogFlow, Azure Bot Service, RingCentral Engage, and Sunshine Conversations allow you to use their APIs to connect with a wide number of voice, chat, messaging, and social channels. Furthermore, the platform you’re looking at should also be able to integrate with any number of different third-party tools and services you might wish to use for things like data management, analytics, and BI (business information).

6. Does your platform support machine learning and rules-based bots?

Most chatbots used today are rules-based, which means they are pre-programmed to recognize a number of intents and provide pre-programmed responses. No AI involved. The second type of bot is a genuine AI agent. Rather than all their responses being pre-scripted, these types of agents are trained, usually on huge data sets, using machine learning and deep learning techniques. If you know you will just need one type of bot, then it might be better to go with a platform that specializes in doing just that. If you prefer the flexibility of being able to adopt either approach then ensure your vendor supports both scripted bots and genuine AI agents.